Das HCC.Forschungsseminar gibt uns die Möglichkeit, einen besseren Überblick über vorhandene Aktivitäten am HCC.lab zu erhalten und uns vor allem damit auseinanderzusetzen. 

Alle Themen die für den HCC Forschungsbereich interessant sind, aber insbesondere zu aktuellen Forschungsthemen und –projekten. Vorträge kommen von Mitgliedern des HCC.lab, von BA/MA Theses Student_innen, aber auch von externen Forscher_innen. 

Dieses Seminar soll eine Atmosphäre für gegenseitiges Lernen bieten. Wir wollen in offener Diskussion Neues lernen oder Vorhandenes in Frage stellen. Es geht dabei nicht nur um Themen, sondern auch Methoden, daher lohnt sich die Teilnahme auch an Themen, die nicht unbedingt im thematischen Fokus stehen.

Grundsätzlich ist das Seminar für alle offen, aber alle Mitglieder des HCC.lab sowie alle Studierenden, die ihre BA/MA Thesis bei uns schreiben, sollten grundsätzlich teilnehmen (bei Nichtteilnahme vorher Bescheid geben).

Studierende des Bachelor- und Masterstudiengangs Informatik.

Wechselnd, abhängig vom Vortragsthema.

Topics

Abstracts

Steffen Pade: Development of a web-based application to enable non-technical medical experts to model Bayesian networks for clinical decision support

With more and more advances in medical research as well as the integration of modern technology to store and analyse patients’ data physicians are facing an enormous increase in available information to be considered when making treatment decisions. Especially illnesses like cancer which are difficult to treat, require careful review of all available patient data as well as the utilisation of just the right health examinations to find out what treatment will lead to the best results. This very often includes physicians from various medical fields per patient. To account for this mass of information and the interdisciplinary work, cancer treatments are discussed in so-called Tumor boards - meetings of all involved experts debating on the right way to help their patients. Aiming at supporting physicians to make faster and better decisions researchers at the University Hospital of Leipzig are using Bayesian networks to model the medical condition of patients. Creating this model putting illnesses, their causes, and suitable therapies in relation is currently a manual procedure involving a modeller and a medical expert. This is an extremely time-consuming and time-inefficient process. Accordingly, this thesis’ goal is to create a software supporting medical experts in modelling the network on their own without the need for additional IT support or a deeper understanding of the Bayesian network or the software. By creating a web-based application that enables physicians to input their knowledge about certain medical fields intuitively whenever they want from wherever they want when being connected to the internet, the modelling process will become much faster and efficient.

Andre Breitenfeld: Link the World - A usability study on annotation-based manual relation extraction

The Web of Data is the vision about a world-wide knowledge base. A distributed knowledge base consisting of structured data that states facts about the world. The atoms of these facts are called entities. Entities can be related in almost any way to state facts. The resulting structure is a network from a large number of relation- ships. However, facts are not written in stone and may change. The network is in a constant state of change due to the exploration and publication of new knowledge sources. These knowledge sources, however, need to be preprocessed to become part of the Web of Data. Preprocessing comprises common activities like the identifica- tion of entities and the extraction of relationships between them. A technological cornerstone towards achieving this vision is the semantic annotation. Its goal is to make knowledge explicit by restructuring unstructured data with semantic informa- tion. Concerning the semantic preprocessing, the focus of this work is on the activity of relation extraction. The overall goal is to propose an new approach that facili- tates the process of manual relation extraction in the area of semantic annotation. To achieve this goal, the solution envisioned includes a novel interaction design. In the course of this work, this interaction design will be designed and implemented in a prototypical way. Finally, the designed solution will be evaluated by an empirical usability study. 

Ming-Tung Hong: Computer-Supported Collaborative Knowledge Construction from Annotations for Sensemaking

As a human we make sense of concepts every day, from things in daily basis to the universe. Making sense of a concept is a complex task when it comes to scholarly research questions. Due to the complexity, scholars annotate on their research documents to reflect their impression and personal opinions of a concept. However, with limited capacity of the cognitive load in scholar’s working memory, important ideas and relevant information can be easily lost or overlooked. In our proposal, we want to extend the memory by providing an appropriate user interface that preserve and expand available information. By allowing scholars to visualize and interact with annotations, a software agent builds up scholars’ mental model and knowledge base from other external resources such as Wikidata and ConceptNet. Software agent tries to find connections in the given context and further recommends new hidden connections to scholars. Empowered by software, scholar’s knowledge could possible be expanded and inspired widely and quickly. The way of how to acquire - or even create - knowledge could be changed and impacted by the new nature of computer-supported sensemaking process.